GAI Insights

Jun 4, 2025

I am an MIT Senior Fellow & Lecturer, 5x-founder & VC investing in AI Follow Author Share Comment Technology and financial advisory services concept. Business teamwork and working on digital laptop ... More computer with advisor showing plan of investment to clients at table office. Digital marketing. getty Knowing about some of the trends inherent in AI adoption right now will help pros to triangulate and remain in the fray when it comes to offering a skill set to the market. Think about it – all of this has happened at roughly the speed of light. How do we adapt? Humans are adaptive, but only to a point. Anyway, there are some guidelines that can come in handy when it comes to exploring new AI projects. Some of them have to do with data handling, since at its core, data is the key commodity. A Group of Experts I recently saw a panel at IIA in April with Jean Olive of John Hancock, John Sviokla of GAI Insights, and MIT CSAIL professor Hari Balakrishnan, where the trio discussed some of these principles and how they work in the industry. One of those ideas that Olive raised early on was a move away from iterative product design, towards a broader scope. Sviokla talked about bad incentive structures and how they deter some innovation, noting an acronym EAT for Educate, Apply and Transform. “What you really need is an operational transformation set of measures, where you add back in research and development, and then in the longer term stuff, you do a return on asset measure,” he said. “If we don't have that kind of transformational incentive system at the top of organizations, we're not going to get there.” MORE FOR YOU Balakrishnan, explaining his role of chair of a board at a firm with 500 employees, brought up three essential categories he feels are vital for AI. The first? Create velocity. The second: increase reliability. As the third point, he talked about promoting creativity with data, and the need to “clean up” data for use. “Everyone thinks it's very easy to do, but there's a lot of constraints and a lot of creativity that's needed to do that,” he said. Synthetic or Real? The panel also talked about the use of synthetic data in AI systems. “The fact that you can improve an analysis by generating synthetic data was just not intuitively, obvious … at all,” Sviokla said. “There's only so much information that can generate synthetic data … How does that work?” The People Part “When you can let people connect with other people, whether it's your patient, whether it's your customer, whether it's your producers, that's to me, (the) real power, is getting that velocity, enabling connectivity in that human factor,” Balakrishnan said. “And I don't think we talk about that enough.” Sviokla shared an anecdote in introducing the term “signature skills.” “I did some research ages ago with a woman named Dorothy Barton over at Harvard Business School, and she had this great phrase called signature skills,” he said. “And she said that when people identify with their signature skills … (about) the way the work is done, not the output of the work, they have a really hard time with technological transformation, whereas if they if they attach their pride or identity, whatever, to the outcome of the work, then it's a lot easier than that.” AI Agents In Play Observing the advent of AI agents into the mix, Olive suggested that the challenge is making the right decisions in deployment. Sviokla agreed that agents are going to be huge, referring to a “silicon hive mind” that informs the industry. “I think we've got a lot of examples and things like program trading, object oriented programming, a bunch of other things that we can learn from,” he said. Balakrishnan noted that his customers have been using agents liberally. All of this helps to get to the heart of what enterprise is doing right now – but what companies are doing right now is not what they will be doing in the future. So keep an eye out for those big changes that accompany where we are at as we watch the world change in the blink of an eye.